近體詩格律診斷專家系統之建立

Building a Chinese Classical Poetry Diagnostic Expert System

林敬斌(lin ching pin)

指導教授 ： 曾憲雄

亞洲大學/資訊學院/資訊科學與應用學系碩士班/碩士(2007年)

摘要

唐代近體詩是中國五千年文化的偉大文學創作，不但講求文字所代表的意 涵，也重視格律的優美與對仗工整。唐代近體詩所使用的文字為古漢語，與現在 人們所使用的白話文差異極大，且漢語文字之中多有一字多韻問題，足以影響判 斷格律的正確性。 學習近體詩不僅須瞭解廣大且艱深的格律知識，也須在有限的文字內表達出 創作者的情感，以及符合嚴謹的格律與對仗要求。但人們在創作詩詞時，常以傳 統判斷格律方法來判斷詩作是否符合近體詩嚴格之規則，這不僅會耗費許多時 間，且可能因創作者自身對於詩詞規則的瞭解不深，不易正確判斷詩作是否符合 格律與對仗，使得許多有意創作詩詞的民眾望之卻步，造成學習詩詞創作時的障 礙。雖然已有許多學者提出協助使用者判斷近體詩格律的相關研究，但對於判斷 對仗及一字多韻仍亟需解決。 為了協助判斷詩作是否符合其要求，我們提出『近體詩格律診斷專家系統』 運用專家系統技術，分析詩作是否符合嚴格的格律與對仗要求，並將診斷結果， 提供給使用者作為創作詩詞時的參考。此包含三個階段：1.知識擷取階段：從相關 詩詞書籍之中擷取近體詩格律與對仗相關知識，並加以歸納出格律與對仗規則類 別，找出格律及對仗的規則狀態並加以驗證。2.知識庫的建置：透過知識擷取將格 律與對仗知識規則，建置格律知識庫及對仗知識庫，再輔以判斷格律及對仗所需 之同義詞詞林詞彙庫、中研院中文八萬目詞詞庫及格律資料庫的相互配合，供推 論引擎進行格律與對仗之推論判斷。3.實驗驗證：將唐代詩作透過此系統進行推 論，將判斷結果與詩詞專家或相關書籍的評論相互驗證，以確認系統判斷的正確 性。

關鍵字

格律、對仗、一字多韻、知識擷取、知識庫、推論引擎。

並列摘要

Tang Poetry was the great literary creation of culture of five thousand years in
China. Not only stressed the meaning that the characters represented, but also pay
attention to grace and antithesis of the rules and forms carefully and neatly. The
characters of the Tang Poetry used archaic Chinese, the difference of the writings in the
vernacular that people use was great with now. There is a problem about the rhyme of
the characters, and it is enough to influence judging the exactness of the rules and
forms.
When creating the poetry it should be understands vast and elusive knowledge, not
only merely express the creator's emotion in the limited characters, but also accord with
the rigorous rules and forms and antithesis requirement. As a result of the rules, forms
and antithesis are required; it makes a lot of people loath to create the poetry, and
shackle the learning passion.
Expert System is the main idea of the thesis. We analyze the poetry that user input,
whether the poetry is accord with the strict rules and forms, then output the comments
that based on the analysis. We propose a Chinese Classical Poetry Diagnostic Scheme
(CCPDS) with 3-stage methodology to diagnose the poetry which goes by the rules,
forms and antithesis requirement: (1) Stage 1: Knowledge Acquisition, we retrieve the
knowledge rules, forms and antithesis from the book; (2) Stage 2: Knowledge-Based
Construction, we input the rules, forms and antithesis operating with inference engine
to construct the Chinese Classical Poetry Diagnostic Expert System; (3) Stage 3:
Experiment Verification, to test the correction of the CCPDS, we input several
poetries, then survey the result of the system output with domain expertise.

參考文獻

[6].許正中，唐代絕句析賞，二版，台北，東大圖書股份有限公司，民國九十一年

[14]. 俞士汶， 胡俊峰， 唐宋詩之詞匯自動分析及應用, LANGUAGE AND

LINGUISTICS 4.3:631-647, 2003

Suggested Upper Merged Ontology”Proceedings of the IEEE International

System,” master thesis, Department of Computer and Information Science,